sched/sched/sched_cpuselect.c:  For SMP, In order to find the cpu with the lowest priority thread, we have to remember the already found lowest priority.  Noted by Anonymous in Issue #75

lpc43xx Ethernet:  Fix some backward logic setting full-duplex and 100mbps when autoconfiguration is disabled.  Noted by Anonymous in Issue #76

arch/arm/include/stm32 and stm32f7:  Remove ltdc.h and dma2d.h.  Those header files in that location permitted inclusion into application space logic and, hence, facilitated and encouraged calling into the OS and violating the portable POSIX OS interface.  The definitions in those header files were move the appropriate location in the counterpart, architecture specific files at arch/arm/src/stm32 and stm32f7 dma2d.h and ltdc.h.

configs/stm32f429i-disco/ltdc:  This configuration has been deleted because it violated the portable POSIX OS interface.  It used apps/examples/ltdc and include ltdc.h and dma2d.h which were also removed for the same reason.

fs/userfs:  Fix return value from dup method.  In order to return a pointer, the parameter must be a pointer to a pointer.

This comment converts the underlying IPC used by the UserFS from Unix domain local sockets to UDP LocalHost loopback sockets.  The problem with the local sockets is that they do require operations on the top level psuedo-file system inode tree.  That tree must be locked during certain traversals such as enumerate mountpoints or enumerating directory entries.

This conversion is unfortunate in the sense that Unix local domain sockets are relatively lightweight.  LocalHost UDP sockets are much heavier weight since they rely on the full UDP stack.  If anyone is up for a complete redesign, then using some shared memory and a POSIX message queue would be lightweight again.

This commit also fixes several bugs that were not testable before the inode tree deadlock.  I cannot say that the logic is 100% stable but it does not have basic functionality.

Squashed commit of the following:

    fs/userfs:  Order locking so that access to the shared I/O buffer is also locked.
    fs/userfs:  Converts to use LocalHost UDP loopback for IPC.

Master

* cosmetics

* stm32_hrtim: add helper macros

* smps: cosmetics

* stm32f33xxx_adc: injected channels support, fix some definitions, add interface to disable interrupts

* stm32f334-dsico: beginning of lower-half driver for SMPS (buck-boost onboard converter)

* nucleo-f334r8/highpri: missing ADC trigger configuration

Approved-by: Gregory Nutt <gnutt@nuttx.org>

stm32f103-minimum: Add board_usbmsc_initialize to stm32f103-minimum

Kconfigs: Add CONFIG_LCD_UPDATE that works like CONFIG_NX_UPDATE but can be enabled without enabling the graphics subsystem.

stm32f334-disco, nucleo-f334r8: add missing ram_vectors configuration in linker script

Approved-by: Gregory Nutt <gnutt@nuttx.org>

This reverts commit 63a8d795.

mac802154: Fixes a warning for unused variable and returns -1 from macnet_notify() if event is not used.

Approved-by: Gregory Nutt <gnutt@nuttx.org>

wireless/ieee802154:  Fix a warning found in build testing.  Mouse input:  Cosmetic changes while reviewing logic.

Another fix from recent PR:  mac802154_primitive_free() does not exist.  Perhaps ieee802154_primitive_free() is what was intended?

drivers/mtd/filemtd.c: Don't use file descriptors... Use the internal file system interfaces so that the loop device can be shared across threads.

drivers/loop: Don't use file descriptors... Use the internal file system interfaces so that the loop device can be shared across threads.

According to the specification, the close function must wait until all data
has been written before it closes the file (except O_NONBLOCK is set). The
maximum waiting time for this is not specified.

To be able to edit the file list of the process, the close function has to lock
the file list semaphore. After that the close function of the serial driver is
called.

Waiting for the complete transmission of all data is done in the serial driver.
This causes the semaphore to remain locked until all data has been sent.
However, no other thread of the process can edit the file list for that time
(open, close, dup2, etc.). This is not optimal in a multithreaded environment.
Therefore, we have to keep the waiting time within the driver as short as possible.

fs/userfs:  There are some deadlock issues that make the UserFS un-usable at the current time.  Added to the TODO list; also feature is now marked EXPERIMENTAL.

This commit modifies the Unix domain local socket design.  Local sockets are built on top of pipes.  The Local socket implementation maintained file descriptors to interrupt with the pipes.  File descriptors have the bad property that they are valid only while running on the thread within the task that created the local socket.

As a policy, all internal OS implementations must use "detached" files which are valid in any context and do not depend on the validity of a file descriptor at any point in time.  This commit converts the usage of file descriptors to detached files throughout the local socket implementation.

Squashed commit of the following:

    net/local: Finish change to eliminate use of file descriptors.
    net/local:  A little more of the conversion.
    net/local: Beginning of chnages to eliminate use of file descriptors in the local socket implementeation. poll() will be a problem.

ieee802154: Simplifies notify() and rxframe() calls to a single notify() call. dataind's and all other "notifs" are now "primitives" which aligns with standard terminology

* mac802154: Adds missing breaks from case statement

* sixlowpan: Fixes bad logic where ACK is not requested if address is not a broadcast

* ieee802154: Simplification of "notifs" and "datainds" to generic primitives passed via a single notify call to the layer above the MAC

* Directories.mk should reference CONFIG_WIRELESS instead of CONFIG_DRIVERS_WIRELESS

* xbee_netdev: Network must be locked when calling sixlowpan_input

* sixlowpan: Reassembly buffer can't be freed if provided by radio driver

* sixlowpan: Don't free IOB if there is an error processing it as the MAC will try to pass it along to another receiver

* ieee802154: Adds basic logging to ieee802154_primitive.c

* Minor fixes after rebase

* xbee: Adds AT query timeout to retry if XBee doesn't respond to request

* same70-xplained: Adds Xbee support. Makes mikroBus slot Kconfig 'choice'

* mac802154: Removes unused function declaration

* drivers/mrf24j40: Fixes compilation error using . operator rather than -> operator

* mac802154_device: Changes a few mac802154_primtive_free's to ieee802154_primitive_free() and changes notif to primitive in a couple places.

* mac802154: Adds promiscous mode logic to bypass parsing of incoming frames. MAC char device also checks for promiscous mode and passes whole frames including header and FCS to the application if promiscous mode is enabled.

* sixlowpan: Fixes logic to correctly check if packet is large enough to include header.  This would cause packets to be considered too small when they are sufficiently sized.

* sixlowpan: Fixes forwarding logic to use forwarding device rather than received device to look up destination link layer address

* net/ipforward: Fixes typo that caused build error when IP forwarding was enabled with CONFIG_NET_ICMPv6_NEIGHBOR enabled as well.

* configs/same70-xplained: Simple spelling fix

Approved-by: Gregory Nutt <gnutt@nuttx.org>

Update TODO for broken local socket feature; Add work around to UserFS for broken local socket feature.

 libc/userfs: Correct return value from dispatchers.  Should return zero on success, not the number of bytes sent.

drivers/wireless/ieee80211:  Fix typos and spelling errors as needed for Photon build.

Fixes a memory leak that is caused because the client message queue is not unlinked after the client disconnects from the NX server.